feat(growth): add openclaw trend x playbook and post-length lint

Rex · Rex · commit fc48bd022633 · 2026-03-05T17:38:01.000+08:00
diff --git a/docs-site/blog/index.md b/docs-site/blog/index.md
@@ -6,6 +6,7 @@ Practical notes for builders choosing an Agent OS and scaling from local prototy
 
 - [What Is LoopForge? (Formerly LoopForge)](what-is-loopforge.md)
 - [LoopForge vs OpenFang/OpenClaw (Builder View)](rexos-vs-openfang-openclaw.md)
+- [OpenClaw Trend: LoopForge Optimization + X Playbook](openclaw-trend-x-playbook.md)
 - [Editorial Calendar](editorial-calendar.md)
 
 ## Suggested reading order
@@ -14,4 +15,5 @@ Practical notes for builders choosing an Agent OS and scaling from local prototy
 2. First successful run: [New User Walkthrough](../tutorials/new-user-walkthrough.md)
 3. Practical templates: [10 Copy/Paste Tasks](../examples/case-tasks/ten-copy-paste-tasks.md)
 4. Positioning and alternatives: [LoopForge vs OpenFang/OpenClaw (Builder View)](rexos-vs-openfang-openclaw.md)
-5. Publishing rhythm: [Editorial Calendar](editorial-calendar.md)
+5. This week's growth playbook: [OpenClaw Trend: LoopForge Optimization + X Playbook](openclaw-trend-x-playbook.md)
+6. Publishing rhythm: [Editorial Calendar](editorial-calendar.md)
diff --git a/docs-site/blog/openclaw-trend-x-playbook.md b/docs-site/blog/openclaw-trend-x-playbook.md
@@ -0,0 +1,60 @@
+# OpenClaw Trend: LoopForge Optimization + X Playbook
+
+Last updated: **March 5, 2026**.
+
+This page turns competitor signals into concrete execution for LoopForge: what to improve next, and what to post on X this week.
+
+## Source anchors
+
+- OpenClaw GitHub README (positioning, onboarding wizard, channel breadth):  
+  https://github.com/openclaw/openclaw
+- OpenClaw docs (getting started / wizard / onboarding):  
+  https://docs.openclaw.ai/start/getting-started  
+  https://docs.openclaw.ai/start/wizard
+- LoopForge CLI and onboarding baseline docs:  
+  ../reference/cli.md  
+  ../explanation/reliability-baseline.md
+
+## What OpenClaw validates in the market
+
+1. Onboarding funnel quality is a growth engine (`openclaw onboard` as default path).
+2. Scenario coverage and links reduce evaluation friction.
+3. Public docs + release rhythm create compounding trust.
+
+Inference: OpenClaw momentum is real, but its center of gravity is personal assistant breadth. LoopForge should use the same growth mechanics while keeping our engineering-delivery positioning.
+
+## LoopForge optimization backlog (P0/P1)
+
+### P0 (ship in 1-2 weeks)
+
+1. Publish onboarding reliability counters weekly.
+   - Artifact: `onboard-report.md` from `scripts/onboard_metrics_report.py`.
+2. Expand copy/paste task packs from "examples" into vertical bundles.
+   - Start with: repo migration, test stabilization, release audit.
+3. Add X-ready snippets for each feature release.
+   - Every release note includes 2 short posts and 1 thread opener.
+
+### P1 (ship in 2-4 weeks)
+
+1. Add "from failure to recovery" walkthroughs (doctor -> fix -> rerun).
+2. Build a small "proof gallery" page with command/output evidence.
+3. Introduce weekly benchmark cards (time-to-first-success, first-task success rate).
+
+## X posting rule for character limits
+
+1. Target length: **220-260** (buffer for edits/mentions).
+2. Hard limit: **280**.
+3. Validate drafts before publishing:
+
+```bash
+python3 scripts/x_post_lint.py --file docs/marketing/openclaw-trend-x-posts.en.txt --limit 280 --warn-at 260
+```
+
+The linter uses URL weight 23 and East Asian width rules for a practical preflight check.
+
+## Ready-to-post copy pack
+
+- English pack: `docs/marketing/openclaw-trend-x-posts.en.txt`
+- Chinese pack: `docs/marketing/openclaw-trend-x-posts.zh-CN.txt`
+
+Use the file blocks split by `---` as one-post units.
diff --git a/docs-site/zh-CN/blog/index.md b/docs-site/zh-CN/blog/index.md
@@ -6,6 +6,7 @@
 
 - [什么是 LoopForge？（原 RexOS）](what-is-loopforge.md)
 - [LoopForge 与 OpenFang/OpenClaw 对比（开发者视角）](rexos-vs-openfang-openclaw.md)
+- [OpenClaw 风口：LoopForge 优化与 X 推广作战页](openclaw-trend-x-playbook.md)
 - [内容发布日历](editorial-calendar.md)
 
 ## 建议阅读顺序
@@ -14,4 +15,5 @@
 2. 再跑通首条链路： [新人复习](../tutorials/new-user-walkthrough.md)
 3. 最后套模板实操： [10 个可复制任务](../examples/case-tasks/ten-copy-paste-tasks.md)
 4. 对比竞品场景： [LoopForge 与 OpenFang/OpenClaw 对比（开发者视角）](rexos-vs-openfang-openclaw.md)
-5. 固定输出节奏： [内容发布日历](editorial-calendar.md)
+5. 本周增长打法： [OpenClaw 风口：LoopForge 优化与 X 推广作战页](openclaw-trend-x-playbook.md)
+6. 固定输出节奏： [内容发布日历](editorial-calendar.md)
diff --git a/docs-site/zh-CN/blog/openclaw-trend-x-playbook.md b/docs-site/zh-CN/blog/openclaw-trend-x-playbook.md
@@ -0,0 +1,60 @@
+# OpenClaw 风口：LoopForge 优化与 X 推广作战页
+
+最后更新：**2026 年 3 月 5 日**。
+
+本文把竞品信号转成可执行动作：LoopForge 该先优化什么，以及这周在 X 上该怎么发。
+
+## 证据来源
+
+- OpenClaw GitHub README（定位、onboard 向导、渠道覆盖）：  
+  https://github.com/openclaw/openclaw
+- OpenClaw 文档（getting started / wizard / onboarding）：  
+  https://docs.openclaw.ai/start/getting-started  
+  https://docs.openclaw.ai/start/wizard
+- LoopForge 当前文档（CLI 能力与可靠性基线）：  
+  ../reference/cli.md  
+  ../explanation/reliability-baseline.md
+
+## OpenClaw 这波热度说明了什么
+
+1. 上手漏斗质量直接决定增长效率（`openclaw onboard` 作为默认入口）。
+2. 场景文档足够多，评估成本就会下降。
+3. 高频更新与文档联动，会持续累积信任。
+
+推断：OpenClaw 的势能是真实存在的，但它主打的是“个人助手覆盖广度”。LoopForge 应该借同样的增长机制，继续强化“工程可复现交付”这条主线。
+
+## LoopForge 优化清单（P0/P1）
+
+### P0（1-2 周内）
+
+1. 固定周更 onboarding 可靠性数据。
+   - 产物：由 `scripts/onboard_metrics_report.py` 生成 `onboard-report.md`。
+2. 把案例任务从“示例集合”升级为“行业任务包”。
+   - 第一批：仓库迁移、测试修复、发布审计。
+3. 每个功能发布都配套 X 素材。
+   - 规则：每次 release note 附 2 条短帖 + 1 条串帖开头。
+
+### P1（2-4 周内）
+
+1. 增加“失败到恢复”的演示路径（doctor -> fix -> rerun）。
+2. 上线一个轻量 proof gallery，展示命令和结果证据。
+3. 固化周报指标卡（首跑耗时、首任务成功率、失败类别 TopN）。
+
+## X 发帖长度规则（防超字数）
+
+1. 推荐范围：**220-260**（留改稿余量）。
+2. 硬上限：**280**。
+3. 发帖前本地检查：
+
+```bash
+python3 scripts/x_post_lint.py --file docs/marketing/openclaw-trend-x-posts.zh-CN.txt --limit 280 --warn-at 260
+```
+
+这个 lint 会按 URL=23 字符和东亚宽字符规则做发布前预检。
+
+## 可直接发布的文案包
+
+- 中文：`docs/marketing/openclaw-trend-x-posts.zh-CN.txt`
+- 英文：`docs/marketing/openclaw-trend-x-posts.en.txt`
+
+两个文件都用 `---` 分隔，每一段就是一条可发布草稿。
diff --git a/docs/marketing/openclaw-trend-x-posts.en.txt b/docs/marketing/openclaw-trend-x-posts.en.txt
@@ -0,0 +1,20 @@
+OpenClaw is pushing personal-assistant momentum hard. We are taking the engineering-delivery lane: local-first, reproducible checkpoints, and auditable runs. Try `loopforge onboard` and ship your first verified task in minutes. #OpenClaw #LoopForge #AgentOS
+---
+If your team asks "Can we replay this result next week?", you need an Agent OS built for delivery, not just chat. LoopForge keeps workspace traces, checkpoint commits, and routing control in one CLI. #AIEngineering #LoopForge
+---
+OpenClaw proves onboarding quality drives growth. We are applying the same lesson to dev workflows: strict doctor checks, onboarding success metrics, and daily reliability reports. #OpenClaw #DevTools
+---
+3 commands to evaluate LoopForge today:
+1) `loopforge init`
+2) `loopforge onboard --workspace lf-demo`
+3) `loopforge agent run --workspace lf-demo --prompt "Create notes/plan.md"`
+---
+OpenClaw owns multi-channel assistant breadth. LoopForge focuses on code delivery depth: harness loop (change -> verify -> checkpoint), local-first providers, and reproducible outputs. Different jobs, clear fit. #AgentOS
+---
+Shipping with AI should be measurable. We now log onboarding outcomes and failure categories by default, then roll up daily reports for reliability baselines. This is the kind of evidence teams can act on. #LoopForge #DevRel
+---
+Trend hook: "assistant experience is hot". Counter-position: "engineering reliability is scarce". If you care about passing verifiers and preserving artifact trails, start from LoopForge, not from a generic chat shell. #OpenClaw #AIInfra
+---
+Before posting to X, lint your copy in repo:
+`python3 scripts/x_post_lint.py --file docs/marketing/openclaw-trend-x-posts.en.txt --limit 280 --warn-at 260`
+No surprises at publish time.
diff --git a/docs/marketing/openclaw-trend-x-posts.zh-CN.txt b/docs/marketing/openclaw-trend-x-posts.zh-CN.txt
@@ -0,0 +1,19 @@
+OpenClaw 把“个人助手”热度拉起来了。我们走另一条路：把 AI 变成可交付工程产能。LoopForge = 本地优先 + 可复现 checkpoint + 可审计运行轨迹。`loopforge onboard` 先跑起来。#OpenClaw #LoopForge #AgentOS
+---
+团队真正关心的不是“会不会聊”，而是“下周还能不能复现”。LoopForge 用 workspace 产物、checkpoint 提交和路由配置，把结果稳定下来。#AI工程化 #LoopForge
+---
+OpenClaw 给我们的启发很直接：上手漏斗做得越好，增长越稳。我们对应动作是：严格 doctor、onboard 成功率指标、失败分类日报。#OpenClaw #DevTools
+---
+今天就能验证的 3 条命令：
+1) `loopforge init`
+2) `loopforge onboard --workspace lf-demo`
+3) `loopforge agent run --workspace lf-demo --prompt "Create notes/plan.md"`
+---
+OpenClaw 的强项是多渠道个人助手覆盖；LoopForge 的强项是工程交付深度：change -> verify -> checkpoint，配合本地优先模型路由。场景不同，选型要准。#AgentOS
+---
+AI 落地要看证据，不看口号。我们默认记录 onboarding 成功/失败与失败类别，再做日聚合报告。这样团队每周都能看到真实改进。#LoopForge #工程效率
+---
+借 OpenClaw 风口的正确姿势：不抄“助手叙事”，而是放大“可复现交付”。如果你在做代码任务闭环，先跑 LoopForge 的 harness 链路。#OpenClaw #AIInfra #LoopForge
+---
+发 X 前先验字数，避免临门改稿：
+`python3 scripts/x_post_lint.py --file docs/marketing/openclaw-trend-x-posts.zh-CN.txt --limit 280 --warn-at 260`
diff --git a/mkdocs.yml b/mkdocs.yml
@@ -116,6 +116,7 @@ plugins:
             Blog: 博客
             "What Is LoopForge?": 什么是 LoopForge？
             LoopForge vs OpenFang/OpenClaw: LoopForge 与 OpenFang/OpenClaw
+            "OpenClaw Trend: LoopForge Optimization + X Playbook": OpenClaw 风口：LoopForge 优化与 X 推广作战页
             Editorial Calendar: 内容发布日历
   - search:
       lang:
@@ -211,6 +212,7 @@ nav:
       - Index: blog/index.md
       - What Is LoopForge?: blog/what-is-loopforge.md
       - LoopForge vs OpenFang/OpenClaw: blog/rexos-vs-openfang-openclaw.md
+      - "OpenClaw Trend: LoopForge Optimization + X Playbook": blog/openclaw-trend-x-playbook.md
       - Editorial Calendar: blog/editorial-calendar.md
   - Reference:
       - CLI: reference/cli.md
diff --git a/scripts/tests/test_x_post_lint.py b/scripts/tests/test_x_post_lint.py
@@ -0,0 +1,46 @@
+import importlib.util
+import tempfile
+import unittest
+from pathlib import Path
+
+
+def load_module():
+    module_path = Path(__file__).resolve().parents[1] / "x_post_lint.py"
+    spec = importlib.util.spec_from_file_location("x_post_lint", module_path)
+    module = importlib.util.module_from_spec(spec)
+    assert spec is not None and spec.loader is not None
+    spec.loader.exec_module(module)
+    return module
+
+
+mod = load_module()
+
+
+class XPostLintTests(unittest.TestCase):
+    def test_split_blocks_uses_separator(self):
+        text = "first post\n---\nsecond post\nline2\n---\nthird post\n"
+        blocks = mod.split_blocks(text)
+        self.assertEqual(blocks, ["first post", "second post\nline2", "third post"])
+
+    def test_weighted_length_counts_urls_and_wide_chars(self):
+        text = "LoopForge 上线 https://example.com/path"
+        length = mod.weighted_length(text, url_weight=23)
+        # "LoopForge " (10) + "上线" (4, wide chars) + " " (1) + URL (23)
+        self.assertEqual(length, 38)
+
+    def test_evaluate_posts_marks_over_limit(self):
+        posts = [("post-1", "a" * 281), ("post-2", "b" * 250)]
+        rows = mod.evaluate_posts(posts=posts, limit=280, warn_at=260, url_weight=23)
+        self.assertEqual(rows[0]["status"], "OVER")
+        self.assertEqual(rows[1]["status"], "OK")
+
+    def test_main_returns_nonzero_when_any_post_is_over_limit(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            path = Path(tmp) / "posts.txt"
+            path.write_text("short\n---\n" + ("x" * 281) + "\n", encoding="utf-8")
+            exit_code = mod.main(["--file", str(path), "--limit", "280", "--warn-at", "260"])
+            self.assertEqual(exit_code, 1)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/scripts/x_post_lint.py b/scripts/x_post_lint.py
